☆ Estimations for MNI coordinates adjusted for these variations areavailable on our website (http://brain.job.affrc.go.jp) together with otherrelated tools and reference data. Upon request, we can add new alternatives,provided their descriptions are clear enough to be reproduced virtually inreference MR images.⁎ Corresponding author. Fax: +81 29 838 7319.E-mail address: [email protected] (I. Dan).
1 The two authors contributed equally to this work.Available online on ScienceDirect (www.sciencedirect.com).
The international 10/20 system has stood as the de-factostandard for electrode placement used in electroencephalography(EEG) for half a century. This system describes head surfacelocations via relative distances between cranial landmarks over thehead surface. The primary purpose of the 10/20 system (Jasper,1958) was to provide a reproducible method for placing a relativelysmall number (typically 21) of EEG electrodes over differentstudies, and there was little need for high spatial resolution andaccurate electrode placement.
With the advent of multi-channel EEG hardware systems andthe concurrent development of topographic methods and tomo-graphic signal source localization methods, there was an increasedneed for extending the 10/20 system to higher density electrodesettings. Therefore, the 10/10 system, an extension to the original10/20 system with a higher channel density of 81, was proposed(Chatrian et al., 1985; see Supplementary material 2 for details).After some arguments on the nomenclature of electrode positions(Nuwer, 1987), its modified form has also been accepted as astandard of the American Clinical Neurophysiology Society(ACNS; former American Electroencephalographic Society; Klemet al., 1999; American Electroencephalographic Society, 1994) andthe International Federation of Clinical Neurophysiology (IFCN;former International Federation of Societies for Electroencephalo-graphy and Clinical Neurophysiology; Nuwer et al., 1998).However, high-end users sought even higher density electrodesettings. 128 channel systems are now a common commercialchoice, and even 256 channel EEG systems are commerciallyavailable (Suarez et al., 2000). Thus, Oostenveld and Praamstra(2001) logically extended the 10/10 system to the 10/5 system,enabling the use of more than 300 electrode locations (320 weredescribed explicitly).
In the meantime, the 10/20 system’s primary use began to shiftfrom simply providing guidance for placing EEG electrodes to beingused for direct positional guidance for newly developing transcranialneuroimaging techniques, near-infrared spectroscopy (NIRS; Oka-moto et al., 2004a,b), and transcranial magnetic stimulation (TMS;Herwig et al., 2003). Use of the 10/20 system allows reproducibleprobe or coil settings on scalps of multiple subjects.
1601V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
Moreover, the 10/20 system serves as the standard craniallandmarks for mediating probabilistic registration (Okamoto et al.,2004a; Okamoto and Dan, 2005; Singh et al., 2005; Tsuzuki et al.,2006). In a series of previous papers, we established a method toprobabilistically register any given scalp position to the correspond-ing scalp or cortical point in standard stereotaxic brain coordinatesystems such as MNI (Montreal Neurological Institute) andTalairach systems without the use of MR images of a subject. Sincethese stereotaxic brain coordinates serve as the common spatialplatform for data presentation of conventional tomographicneuroimaging techniques including fMRI and PET (Collins et al.,1994; Talairach and Tournoux, 1988; reviewed in Brett et al., 2002),the registration of stand-alone multi-subject fNIRS and TMS data toa brain template in the MNI standard coordinate system facilitatesboth intra- and inter-modal data sharing within the neuroimagingcommunity. Therefore, the 10/20 system has been gainingimportance as a standard relative head-surface-based positioningmethod for various transcranial brain mapping methods.
However, it is also true that the original 10/20 system has notbeen equipped as a versatile system to fully support such unexpectedapplications. In the process of developing high density settings, the10/20-derived systems have been mainly appreciated as methods toincrease spatial resolution for EEG studies, where more denselypositioned electrodes are proven to be effective in increasing thespatial resolution when the three-dimensional signal sourceestimation is applied (Pascual-Marqui et al., 2002). Meanwhile itsaspect as a relative head-surface-based positioning system has notbeen examined well. In particular, how effectively high-resolutionderivatives of the 10/20 system can separate each cranial landmark,which is especially important for head-surface-based positionalestimation in TMS and NIRS, still remains unknown. Therefore, wewill evaluate the effective spatial resolution of the 10/20, 10/10, and10/5 systems for multi-subject studies. We will focus on two sourcesof variability. First, definitions of landmark placement in the original10/20 system by Jasper (1958) are ambiguous, and this results indifferent interpretations among experimenters and variability amongstudies. Second, even if a fixed definition of landmark placement isused, scalp and cortical anatomies are different among subjects andthis results in inter-subject variability.
To evaluate variability, we performed virtual 10/20, 10/10, and10/5 measurements on MR images that we described previously.Subsequently, we transformed all the scalp data to MNI space andstatistically assessed the spatial variability. In so doing, we soughtto assess the potential of 10/20, 10/10, and 10/5 systems as relativehead-surface-based positioning systems.
Analysis
Unambiguously illustrated 10/10 system
Currently, there are several different branches and derivatives ofthe 10/20 system, which tend to be used without clear definitions.Comparing different derivatives is something of a paradox: there isno unambiguous standard system, yet we must deal with thevariability of the derivatives. As a practical compromise, we willfirst present the “unambiguously illustrated (UI) 10/10 system” asan unambiguous standard. This is not a new invention of ours,rather we simply eradicated ambiguity in the original descriptionand complemented the 10/10 system that was proposed by ACNS(Klem et al., 1999), which is highly compatible with the oneproposed by IFCN (Nuwer et al., 1998).
Here we will present a sufficiently unambiguous description forsetting UI 10/10 positions and add detailed descriptions and relatedissues later, in appropriate contexts. We begin with setting fourdistinct primary reference points on the scalp anatomy: nasion(Nz), a dent at the upper root of the nose bridge; inion (Iz), anexternal occipital protuberance; left preauricular point (LPA), ananterior root of the center of the peak region of the tragus; and rightpreauricular point (RPA) determined as for the left (Fig. 1a). In theUI 10/10 system, LPA and RPA are the same as T9 and T10.
Next, we move on to setting reference curves on a scalp. Forclarity, we define the term “reference curve” as a path ofintersection between the head surface and a plane defined by threegiven points. First, we tentatively set the sagittal central referencecurve using Nz and Iz, with Cz being temporarily defined as theirmidpoint, along the head surface (Fig. 1b). Second, we set thecoronal central reference curve along LPA, Cz, and RPA byadjusting the sagittal central reference curve so that Czequidistantly divides both the sagittal and coronal central referencecurves (Fig. 1c). The sagittal central reference curve thusdetermined in the UI 10/10 system is divided from Nz to Iz, in10% increments to generate Fpz, AFz, Fz, FCz, Cz, Cpz, Pz, POz,and Oz (Fig. 1d). Furthermore, the thus determined coronal centralreference curve is divided in 10% increments from LPA to RPA inorder to generate T7, C5, C3, C1, Cz, C2, C4, C6, and T8 (Fig. 1e).
Then, we set a left 10% axial reference curve along Fpz, T7, andOz (Fig. 1f). For the left anterior quarter, we divide this portion ofthe curve by one fifth increments, from Fpz to T7, to set Fp1, AF7,F7, and FT7 (Fig. 1h). For the left posterior quarter, we divide byone fifth increments, from T7 to Oz, to set TP7, P7, PO7, and O1(Fig. 1i). We do the same for the right hemisphere (Fig. 1g).
Next, we set six coronal reference curves. Since anterior–frontal(AF) and posterior–occipital (PO) reference curves follow slightlydifferent rules, we first deal with four coronal reference curves inthe middle, taking the frontal (F) coronal reference curve as anexample. We define the F coronal reference curve using F7, Fz, andF8 (Fig. 1j). We divide the F7–Fz portion of the curve by onefourth increments, from F7 to Fz, to generate F5, F3, and F1 (Fig.1k). We do the same for the F8–Fz portion on the right hemisphere(Fig. 1l). We apply the same quarterly division rule on eachhemisphere to the fronto-central/temporal (FC/FT), temporo-/centro-parietal (TP/CP), and parietal (P) coronal reference curves(Fig. 1m). Next, we determine the anterior–frontal (AF) coronalreference curve using AF7, AFz, and AF8 (Fig. 1n). Sincequarterly division results in overcrowded positions, the AF7–AFzportion of the curve is only bisected to generate AF3, and the AFz–AF8 portion, to generate AF4. Similarly, we work on the parieto-occipital (PO) coronal reference curve to set PO3 and PO4.
Finally, we set a left 0% axial reference curve along Nz, LPA(T9), and Iz (Fig. 1o). For the left anterior quarter, we divide by onefifth increments, from LPA (T9) to Nz, to set FT9, F9, AF9, and N1(Fig. 1p; Klem et al., 1999). For the left posterior quarter, we divideby one fifth increments, from LPA (T9) to Iz, to set TP9, P9, PO9,and I1 (Fig. 1q). We do likewise for the right hemisphere to set N2,AF10, F10, FT10, TP10, P10, PO10, and I2.
For EEG studies, A1 and A2 electrodes are placed on the leftand right ear lobes, but they are not important for other transcranialmodalities. To maintain inter-modal generality, we do not deal withA1 and A2 in the current study.
In this way, we determined 81 positions (excluding A1 and A2)of the UI 10/10 systems with our best effort to exclude anyambiguity (Fig. 1r; circled points in Fig. 6 and in Supplementary
1602 V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
material 1). These 81 positions include all the positions described inthe 10/10 system proposed byACNS.We use the UI 10/10 system asthe standard for inter-system comparison in the rest of the study.
The UI 10/10 system basically provides backward compatibilityto the 10/20 system by Jasper (1958) except for the followingdifferences. There are minor changes in nomenclature: T3, T4, T5,and T6 in the 10/20 system are renamed T7, T8, P7, and P8respectively in the 10/10 system (Klem et al., 1999). The 10/20system includes Pg1 and Pg2 placed on the pharynges. Since theyare extremely hard to describe statistically, we excluded them fromthe analysis. The original 10/20 system by Jasper (1958) includesCb1 and Cb2, which are supposed to be on the scalp above thecerebellum. To maintain backward compatibility, we excludedthem from the UI 10/20 system. Thus, we defined 19 positions inthe UI 10/20 system, excluding A1, A2, Pg1, Pg2, Cb1, and Cb2(black circles in Fig. 6 and in Supplementary material 1).
Terminology regarding 10/20-derived systems
Currently, there are several different branches and derivatives ofthe 10/20 system, which tend to be used without clear definitions. Inorder to avoid any confusion in this study, we will clarify theterminology regarding the major variations of the 10/20 system. Thesources that we used in our determination are summarized in Table 1.By “Jasper’s 10/20” system, we refer to the 10/20 system that wasdescribed by Jasper (1958).We refer to the original ACNS definitionas the “ACNS 10/10 system”. We call the 10/10 system proposed bythe IFCN the “IFCN 10/10 system” (Nuwer et al., 1998). It isbasically the same as the ACNS 10/10 system except for thefollowing differences: the IFCN 10/10 system prefers Jasper’soriginal nomenclature, T3, T5, T4, and T6, rather than T7, P7, T8,and P8 as in the ACNS 10/10 system; the IFCN 10/10 systemdescribes only 10/10 positions on or above the 10% axial referencecurves albeit it does not exclude the possibility of using 10/10positions on the 0% axial reference curve. When we do not have todistinguish between them, we call them collectively the “ACNS/IFCN 10/10 system”. Chatrian et al. (1985) was the first to describethe 10/10 system, but we refer to it as “Chatrian’s 10/10” system astheir method is slightly different from the ACNS/IFCN andUI 10/10systems (see Supplementary material 2 for detailed description). Asdescribed in the previous section, we define the “UI 10/10 system”
as an unambiguous complementation for the 10/10 system proposedby ACNS (Klem et al., 1999). In addition, as described above, the UI10/20 system was defined so that all the positions were included in
Table 1The number of standard positions in various 10/20-derived systems
System Number of standard positions Additional implicated
Jasper's 10/20 19 (25 if A1, A2, Cb1, Cb2,Pg1, and Pg2 are included)
Nz, Iz, right and left pFpz, Oz, C5, C6, (A1
Chatrian's 10/10 81 Right and left preauriIFCN 10/10 64 Iz, right and left preau
positions on and beloACNS 10/10 75Oostenveld's 10/20 21Oostenveld's 10/10 85Oostenveld's 10/5 320 Nine positions that mUI 10/20 19UI 10/10 81UI 10/5 329
the UI 10/10 system. Oostenveld and Praamstra (2001) proposed the10/5 system, which we refer to as “Oostenveld’s 10/5” system, butwe also use the term “OostenveldTs 10/10” system when specificallyselecting 10/10 positions from among Oostenveld’s 10/5 positions.Other variations will be discussed below.
Subjects and data analysis
We reanalyzed the MRI data sets of the 17 healthy volunteers(mongoloid; 9 males, 8 females; aged 22 to 51 years) withinformed consent, which we had subsequently registered in theMNI coordinate system in a previous study (Okamoto et al.,2004a). Detailed methods for image processing, transformation tothe MNI space, and virtual-head-surface landmark measurementswere as previously described (Jurcak et al., 2005).
Briefly, we extracted head and brain images from theMRI data setsof the 17 subjects to produce isotropic images of 1×1×1 mm voxelsin size containing 8-bit continuous-tone data. These were subse-quently converted to 2-bit data. The 10/20, 10/10, and 10/5 positionswere determined according to the distance between landmarks overthe head surface. Basically, we calculated the distance between a set ofpoints over the head surface in a virtual space by defining a planeusing three landmark positions. We extracted head surface pointswhich comprised a cross-section between the plane and the headsurface and drew a reference curve utilizing the extracted points.When only two points are given, we used the shortest distant searchalgorithm. This sets numerous planes intersecting the two points.Among the cross-sections between the planes and the head surface,wechose the one that gave rise to the shortest path along the head surface.
After multi-subject data for a given landmark position wasexpressed in MNI space, we calculated the mean coordinatelocations across subjects as,
where n is the number of subjects, and x, y, and z are MNIcoordinate values for a given landmark point of a subject. The
positions The most reliable source
reauricular points,, A2, Cb1, Cb2, Pg1, Pg2)
Fig. 6 in Jasper (1958)
cular points Figs. 1 and 2 in Chatrian et al. (1988)ricular points, electrodew 0% axial reference curve
Fig. 1 in Nuwer et al. (1998)
Fig. 7 in Klem et al. (1999)Fig. 1 in Oostenveld and Praamstra (2001)Fig. 1 in Oostenveld and Praamstra (2001)
ay interfere with the eyes Fig. 2 in Oostenveld and Praamstra (2001)Fig. 6 of this manuscriptFig. 6 of this manuscriptFig. 6 of this manuscript
1603V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
mean coordinate location provides the most likely estimates of thegiven point. Meanwhile, SD provides the measure for its variabilityacross subjects within a given system and can be called inter-subject or intra-system variability, depending on the context.
Primary reference points
The original 10/20 system, at the time of its invention, wasprimarily intended for placing a relatively small number of elec-trodes in a balanced, reproducible manner over the scalp (Jasper,1958). The description was only fine enough to support sparseelectrode placement with accuracy in the order of centimeters.
10/20 measurement starts with setting four distinct primaryreference points on the scalp anatomy, but the definitions ofthese reference points themselves are somewhat ambiguous.Among them, nasion, a dent at the upper root of the nosebridge, is the clearest and can easily be detected precisely. Inion,an external occipital protuberance, is less visible. Even if thestructure is distinct enough, it is felt as a patch with a diameterof only several millimeters. For some subjects, the structure isoften undetectable and an experimenter has to estimate itslocation from neighboring anatomical structures such as trapeziusmuscles.
Preauricular points are also a source of ambiguity. Accordingto Jasper’s original description, they were defined as depressionsat the root of the zygoma just anterior to the tragus (Jasper,1958). However, it is difficult to pinpoint the root of the zygomaat the skin, and the size of the tragus is approximately 1 cm.These factors make a precise, reproducible detection of thepreauricular points difficult. Some laboratories have resolved thisambiguity with minor adjustments of their own. For example,when MR images of a subject are available, an external ear canalprovides a stable anatomical guide. Another popular modificationis the center of the peak region of the tragus, which is obviousfor most subjects. As far as we know, the most stable localdefinition seems to be the dent between the upper edge of thetragus and the daith, which can be identified as a small point.Thus, for the preauricular points, we used four differentdefinitions: the upper limit of the external ear canal, the centerof the peak region of the tragus, the dent between the upper edgeof the tragus and the daith, and, as a minor modification ofJasper’s rather ambiguous definition, we tentatively restricted thepreauricular region to the point located at the anterior root of thecenter of the peak region of the tragus, which can be detected inMR images. We used the same definition for the UI 10/10system. In the rest of the current study, we will use this definitionof preauricular points unless stated otherwise.
We first tested the inter-subject variability of these referencepoints by transferring them onto the MNI space and performing agroup analysis. It should be noted that the variability includes thefollowing two inseparable error sources: the structural differencesof the external landmarks among individuals, and human errorwhen detecting them manually. As Fig. 2 shows, the location of thenasion was almost invariant, while that of the inion had a largerstandard deviation. These variabilities are intrinsic limitations inthe accuracy of 10/20-derived systems and inevitable. The fourlocal definitions of preauricular points lead to slightly differentlocations, but their precision in terms of standard deviations wasalmost the same.
Difference in the preauricular locations also affected thelocations of other UI 10/10 positions. As Fig. 3 shows, temporal
10/10 positions near ears were largely affected by differences in thepreauricular definitions. Deviation decreased in anterior, posterior,and parietal directions so that the 10/10 positions on prefrontal,parietal, and occipital regions were almost unaffected.
Sagittal central reference curve
After the four primary reference points (the nasion, the inion,and the preauricular points) are set, the sagittal central referencecurve between Nz and Iz is determined. The original description ofthe 10/20 system proceeds to the determination of sagittal centralreference curve immediately after primary reference pointdetermination (Jasper, 1958). This process is equivalent to defininga curve in a three-dimensional area with only two points, andtheoretically is not valid. One more rule is necessary. Therefore,Oostenveld and Praamstra (2001) suggested a readjustment of thetentatively drawn sagittal central reference curve so that Cz, or apoint equidistant from both Nz and Iz, is also located at a pointequidistant from the two preauricular points in order to maintaininter-hemispheric balance. The addition of Cz to Iz and Nz enabledthe formation of a plane, and its intersection with the scalp yieldeda unique sagittal central reference curve. Our virtual 10/20determination program adopted the same adjustment procedure(Jurcak et al., 2005). The ACNS/IFCN 10/10 system inherits thesagittal central reference curve definition ambiguity of Jasper’s 10/20 system (Jasper, 1958; Klem et al., 1999; Nuwer et al., 1998),but in the UI 10/10 system used in this study, we applied inter-hemispheric balancing. In all branches of 10/10 systems, thesagittal central reference curve thus determined is divided from Nzto Iz in 10% increments to generate Fpz, AFz, Fz, FCz, Cz, Cpz,Pz, POz, and Oz.
Coronal central reference curve
In JasperTs definition, the coronal central reference curve is onthe plane made by the preauricular points and Cz (Jasper, 1958).The ACNS 10/10 system divides the coronal central referencecurve in 10% increments from the left preauricular point (T9) to theright (T10) in order to generate T7, C5, C3, C1, Cz, C2, C4, C6,and T8, and thus we implemented the same strategy in the UI 10/10system.
In contrast, Oostenveld and Praamstra (2001) used a differentprocess. After determining the preauricular points (labeled LPAand RPA, equivalent to T9 and T10 respectively in the UI 10/10system), an initial central coronal curve LPA–Cz–RPA is used toset points 10% above the preauricular points. In the current study,we tentatively designate them as pseudo-T7 and pseudo-T8(equivalent to T7 and T8 in the UI 10/10 system). On the lefthemisphere, a 10% horizontal reference curve was determined by aplane defined by Fpz, pseudo-T7, and Oz. Along this horizontalcurve, T7 was set as a point equidistant from Fpz and Oz. Thesame process was applied to the right hemisphere to determine T8.Next, a coronal central reference curve was defined by a planedefined by T7, Cz, and T8 (Oostenveld’s T7 and T8 are notnecessarily identical to the UI T7 and T8, i.e., Oostenveld’spseudo-T7 and -T8). Finally, the contour was bisected in halvesand points C5, C3, C1, and C6, C4, C2 were determined on eachhemisphere at the quarter points of T7–Cz and T8–Cz respectively.
This method may not seem as straightforward as the UI method,but it stands on a carefully considered practical compromise.Jasper’s 10/20, ACNS/IFCN 10/10, and the UI 10/10 systems are
Fig . 1.
Fig . 2.
1604 V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
mainly dependent on the central sagittal, central coronal, and 10%axial reference curves. However, we can only choose two referencecurves to achieve equidistance divisions because the scalp is notsymmetric. Though not mentioned explicitly, Jasper’s 10/20 andACNS/IFCN 10/10 systems divide the sagittal and coronal centralreference curves at 50%. This inevitably results in an unbalanceddivision of the 10% axial reference curve, namely the distancesFpz-T7, Fpz-T8, Oz-T7, and Oz-T8 may differ. In contrast,Oostenveld’s 10/10 system divides the sagittal and 10% axialreference curves at 50%. This results in equality in the lengthsbetween Iz–Cz and Nz–Cz, Fpz–T7 and Oz–T7, and Fpz–T8 andOz–T8, while the lengths between T7–Cz and T8–Cz may differ.However, the angles formed by T7, Cz, and pseudo-T7, and T8,Cz, and pseudo-T8 are small. Given that Cz to pseudo-T7 and Czto pseudo-T8 distances are the same, Cz to T7 and Cz to T8distances are also virtually the same. Therefore, Oostenveld’s 10/10 system can place the landmark points on the three referencecurves in a more balanced manner than the UI 10/10 system.
However, since the stability of Oostenveld’s 10/10 system hasonly been implied, we evaluated it using virtual scalp measure-
Table 2Length differences of reference curves between Oostenveld's and the UI10/10 definitions
Values are presented in percentages with standard deviations (if applicable).
1605V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
ments on the MR images of 17 subjects. As shown in Table 2,Oostenveld’s system realized well-balanced divisions of the centralcoronal reference curve with a standard deviation six times lowerthan those of axial divisions of the UI system.
Moreover, we examined the locations of coronal central pointsdetermined by the UI and Oostenveld’s 10/10 systems and theirsubsequent influences on other 10/10 positions. The two methodsreturned nearly identical results for all 10/10 positions (Supple-mentary material 3). As predicted theoretically, intra-systemvariability associated with the estimation was slightly lower inOostenveld’s 10/10 system, reflecting its higher precision.However, the differences are small enough to be negligible inmost practical situations.
We also explored Oostenveld’s 10/10 system’s toleranceregarding the selection of preauricular reference points, which
Fig. 1. Landmark setting procedures for the UI 10/10 system. (a) Primary referencpreauricular points (LPA/T9 and RPA/T10). (b) Sagittal reference curve setting. Wethe setting is only tentative. (c) Central coronal reference curve setting and Cz adjuCz so that it bisects both sagittal and central coronal reference curves. (d) Determina10% increments. (e) Determination of the landmarks on the central coronal referen10% axial reference curve on the left hemisphere. It is set so that Fpz, T7, and Oz alihemisphere. It is set so that Fpz, T8, and Oz align on a plane. (h) Determination oflandmarks are set at one fifth increments. The right anterior portion (not shown) is s10% axial reference curve. The landmarks are set at one fifth increments. The rightThe frontal (F) coronal reference curve is shown. We draw it so that F7, Fz, and F8 athe coronal reference curve. The F coronal reference curve is shown. The landmarkright half portion of the coronal reference curve. The F coronal reference curve is sthe landmarks on the coronal reference curves. As in the F coronal reference curparietal (P) coronal reference curves and the landmarks on them are set. (n) Detreference curves and landmarks. For each hemisphere, AF and PO reference curvereference curve on the left hemisphere. It is set so that Nz, LPA (T9), and Iz align onlandmarks on the left anterior portion of the 0% axial reference curve. The landmasimilar. (q) Determination of the landmarks on the left posterior portion of the 0% aanterior portion (not shown) is similar. (r) All UI 10/10 points are set. The right h
Fig. 2. Locations of primary reference points. All positions are overlaid on the normalso shown. L and R represent left and right, respectively. The centers of the circreference points. The edges represent the boundaries defined by standard deviationas: blue for the anterior root of the tragus, yellow for the peak region of the tragus,upper edge of the tragus and the daith. (A) Frontal view. (B) Occipital view. (C) Lpreauricular points used in this study: the anterior root of the tragus (blue dot), the p(black dot), and the point determined between upper edge of the tragus and the da
are only used tentatively to set the amplitude of T7 and T8.Preauricular points defined as the upper limit of an external earcanal, the center region of the tragus, and the anterior root of thetragus returned almost identical results for all 10/10 positions(Supplementary material 4). The dent between the upper edge ofthe tragus and the daith resulted in a slight upward shift. Thus,Oostenveld’s 10/10 system is tolerant of variation in preauricularpoints in horizontal directions. When all of this is taken together,we demonstrated that Oostenveld’s 10/10 system is optimized forsetting a stable central coronal reference curve.
10% axial reference curve
Setting the 10% axial reference curve can also affect thelocation of 10/20, 10/10, and 10/5 positions. Theoretically, thereare two ways of drawing two axial reference curves: working onanterior and posterior halves or on each hemisphere. In the originaldescription of the 10/20 system by Jasper (1958), hemisphericdivision was implied. Oostenveld’s 10/5 system clearly discussedand applied hemispheric division. To maintain backward compat-ibility with Jasper’s 10/20 system, we adopted hemisphericdivision in the UI 10/10 system in this study.
On the other hand, Le et al. (1998) used anterior/posteriordivision to develop an automatic 10/20 guidance method. We alsointroduced anterior/posterior division primarily because theprobabilistic registration methods we developed do not necessarilyuse all of the 10/20 reference points, hence often those on theanterior or posterior half were sufficient (Jurcak et al., 2005; Singhet al., 2005).
In hemispheric division, the 10% axial reference curve on theleft hemisphere is defined by Fpz, T7, and Oz, and that on the rightby Fpz, T8, and Oz. This leads to Fpz, Fp1, AF7, F7, FT7, T7,TP7, P7, PO7, O1, and Opz aligning on the same plane of the lefthemisphere, and Fpz, Fp2, AF8, F8, FT8, T8, TP8, P8, PO8, O2,and Opz on that of the right. In anterior/posterior division, the 10%
e points. We set four primary reference points: inion (Iz), nasion (Nz), anddraw the sagittal central reference curve and bisect it to set Cz. At this stage,stment. We draw the central coronal reference cure and adjust the location oftion of the landmarks on the sagittal reference curve. The landmarks are set atce curve. The landmarks are set at 10% increments. (f) Determination of thegn on a plane. (g) Determination of the 10% axial reference curve on the rightthe landmarks on left anterior portion of the 10% axial reference curve. Theimilar. (i) Determination of the landmarks on the left posterior portion of theposterior portion (not shown) is similar. (j) Coronal reference curve setting.lign on a plane. (k) Determination of the landmarks on the left half portion ofs are set at one fourth increments. (l) Determination of the landmarks on thehown. The landmarks are set at one fourth increments. (m) Determination ofve, fronto-central/temporal (FC/FT), temporo-/centro-parietal (TP/CP), andermination of the anterior–frontal (AF) and parieto-occipital (PO) coronals are only bisected to produce landmarks. (o) Determination of the 0% axiala plane. The right hemisphere (not shown) is similar. (p) Determination of therks are set at one fifth increments. The right anterior portion (not shown) isxial reference curve. The landmarks are set at one fifth increments. The rightemisphere (not shown) is similar.
alized and averaged head surface images of 17 subjects. MNI coordinates areles represent the most likely locations of MNI coordinates for the primary. Nasion and inion are shown as green circles. Preauricular points are shownblack for the external ear canal, and red for the point determined between theeft temporal view. (D) Schematic of an ear showing the four definitions ofeak region of the tragus (yellow dot), the upper limit of the external ear canalith (red dot).
1606 V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
axial reference curve is defined by T7, Fpz, and T8 on the anteriorhalf, and T7, Oz, and T8 on the posterior half. This leads to T7,Ft7, F7, AF7, Fp1, Fpz, Fp2, AF8, F8, FT8, and T8 aligning on thesame plane on the anterior half, and T7, TP7, P7, PO7, O1, OpzO2, PO8, P8, TP8, and T8 on the posterior half.
It should be noted that Jasper’s 10/20 system was furtherambiguous in setting the reference points on the 10% axialreference curve (Jasper, 1958). From Fpz to Oz, Jasper suggestedsetting the reference points at 10, 30, 50, 70, and 90% distances,but since Fpz–T7 and T7–Oz distances could not be equal, this wastheoretically impossible. Consequently, researchers should eitherneglect small differences for rough measurements or work on eachquarter. To avoid such inconsistency, we chose to work on eachquarter in the UI 10/10 system, and also in the anterior/posteriordivision used in this study. Namely, each Fpz–T7, T7–Oz, Oz–T8,and T8–Fpz portion of 10% axial reference curves was divided inone fifth increments to generate 10/10 positions.
We examined how hemispheric and anterior/posterior divisionsaffect the locations of reference points on 10% axial referencecurves. As Fig. 4 shows, the hemispheric division resulted inslightly lower 10/10 positions on lower parts of the scalp especiallythose on 10% and 0% axial reference curves. Thus, there are non-negligible inter-system differences between 10/10 positionsdetermined by hemispheric and anterior/posterior divisions.However, intra-system variability associated with the estimationwas at similar levels.
Coronal reference curves
In the original description of the 10/20 system by Jasper(1958), mid-frontal and mid-parietal positions (F3, F4, P3, and P4)seemed, according to our best estimation, to be located on thecoronal reference curve along planes defined by F7, Fz, and F8,and P7, Pz, and P8. Therefore, in the UI 10/10 system, we setcoronal reference curves along planes defined by three points.However, this definition is difficult to realize practically withoutextreme precision when measuring the scalp. As far as we haveobserved, experimenters tend to locate mid-frontal and mid-parietal positions slightly below the exact coronal curve. Thus, inour former study, we mimicked this human measurement by
Fig. 5. Effects of the selection of coronal curves on the location of 10/10 positions. Btemplate are the same as in Fig. 3. Blue circles represent points lying on the planesagittal reference curve, as determined according to the UI 10/10 system. Red circlcorresponding points on the 10% axial reference curve and the sagittal central refer10% axial reference curves that are not affected by the selection. The asterisks andear lobes, respectively. (A) Frontal view. (B) Occipital view. (C) Left temporal vie
Fig. 4. Effects of the selection of 10% axial reference curves on the location ocoordinates, and a scalp template are the same as in Fig. 3. Blue circles representhemisphere defined by planes made by Fpz–T7–Oz and Fpz–T8–Oz points, accoraxial reference curves were determined on the anterior and posterior half defined bynot theoretically affected by the 10% axial reference curve selections are shown in pand on or beneath the ear lobes, respectively. (A) Frontal view. (B) Occipital view
Fig. 3. Effects of primary reference point selection on the location of the UI 10/10surface images of 17 subjects. MNI coordinates are also shown. The centers of tstandard positions. The edges represent the boundaries defined by standard deviashown as: blue for the anterior root of the tragus; yellow for the peak region of thethe upper edge of the tragus and the daith. Color for 10/10 locations depends on taffected are shown in pink. The asterisks indicate 10/10 positions that can be on t(i.e., TP9 and TP10). (A) Frontal view. (B) Occipital view. (C) Left temporal vie
setting F3 at the middle of the shortest-distance curve between F7and Fz (Jurcak et al., 2005). This tendency seems more obviouswhen setting AF7–AFz–AF8 and PO7–POz–PO8 coronal refer-ence curves for the 10/10 system.
Fig. 5 shows 10/10 standard positions defined by the twodefinitions of the coronal reference curves described above. Intra-system variability was bigger when coronal curves were defined bythe shortest-distant method. There was also an obvious shift wherethe shortest-distant method tended to locate anterior points furtheranterior, and posterior points further posterior.
Oostenveld’s 10/5 system
As demonstrated above, provided that the choices of the primaryreference points and reference curves are clear enough, 10/20 and10/10 systems can set precise and reproducible scalp landmarks. Thenext question is how far the system can be extended. To date, a 10/5system with more than 300 distinct scalp landmarks has beenproposed (Oostenveld and Praamstra, 2001), but there is thepossibility of excessive landmark setting. Therefore, we examinedthe variability of 10/5 positions for multi-subject studies byperforming virtual 10/5 measurements on the MR images of 17subjects. The measurements were kept as close as possible to theoriginal description by Oostenveld and Praamstra (2001).
Nomenclature for the 10/5 standard positions is presented inSupplementary material 5. Briefly, the central sagittal, coronal, and10% axial reference curves were drawn as in Oostenveld’s 10/10system, described above. On the central sagittal and 10% axialreference curves, 10/5 standard positions were set at 5% intervals.Exceptions to this rule were the anterior and posterior 5% points on10% axial reference curves, which Oostenveld and Praamstra didnot describe clearly. We designated these points as Fp1h, Fp2h,O1h, and O2h by extending their logic on the 10% axial curve. Thepoint between Nz and Fpz was designated as NFpz and nearbypoints on the 5% axial curve designated as NFp1h and NFp2h. Onthe central coronal curve on the T7–Cz–T8 plane, we worked oneach hemisphere separately setting 10/5 standard positions at12.5% intervals between T7 and Cz and repeating the sameprocedure on the right hemisphere. For other coronal referencecurves posterior to AFp and anterior to POO, we used the same
asic descriptions of circles showing 10/10 positions, coordinates, and a scalpdefined by two starting points on the 10% axial curve and one point on thees represent points that lie on curves with the shortest distance between twoence curve. Pink circles represent points on the central coronal, sagittal, andsharps indicate 10/10 positions that can be on the eyes, and on or beneath thew. (D) Right temporal view. (E) Top view.
f 10/10 positions. Basic descriptions of circles showing 10/10 positions,10/10 positions after 10% axial reference curves were determined on eachding to the UI 10/10 system. Red circles represent 10/10 positions after 10%planes made by T7–Fpz–T8 and T7–Oz–T8 points. The UI positions that areink. The asterisks and sharps indicate 10/10 positions that can be on the eyes,. (C) Left temporal view. (D) Right temporal view. (E) Top view.
positions. All positions were overlaid on the normalized and averaged headhe circles represent the most likely locations of MNI coordinates for 10/20tion. Nasion and inion are shown as green circles. Preauricular points aretragus; black for the external ear canal; red for the point determined betweenhe preauricular point selection. The UI 10/10 positions that are not virtuallyhe eyes (i.e., N1, N2), and sharps indicate those on or beneath the ear lobesw. (D) Right temporal view. (E) Top view.
Fig . 3.Fig . 5.
Fig . 4.
1607V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
strategy as with the central coronal curve. AFp and POO coronalcurves were drawn, but halfway points were omitted.
To our knowledge, there is no clear description for settingpoints on 0% and 5% axial reference curves. There were twopossibilities: extending coronal reference curves dorsally or settingthe 0% and 5% axial reference curves independently. Tentatively,we chose the latter strategy. We set a plane using Nz, LPA, and Izon the left hemisphere, and 10/5 standard positions at 5%distances. However, anterior 5, 10, and 15% points are likely onthe eyes. For the 5% axial curve, we set pseudo-T9h and pseudo-T10h 5% above the preauricular points from the distance LPA–Cz–RPA. Using NFpz, pseudo-T9h, and OIz, we set a plane and 10/5standard positions at 5% distances. Again, anterior 10, 15, and20% points are likely on the eye. We followed the same procedurewith the right hemisphere. In this way, we estimated a total of 329Oostenveld’s 10/5 standard positions (including nine additionalpositions) and described their distribution on the MNI space(Supplementary material 6).
Unambiguously illustrated 10/5 system
Aside from some ambiguity, as mentioned above, we regardOostenveld’s 10/5 system as a well-designed derivative of the 10/20 system with emphasis on the balanced setting of central sagittal,central coronal, and 10% axial reference curves. However, it ispossible in practical situations that researchers adopt thenomenclature of Oostenveld’s 10/5 system but simply extend theACNS/IFCN 10/10 system to the 10/5 system without readjustingthe central coronal reference curve. Therefore, we will extend the
1608 V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
UI 10/10 system to cover 10/5 points and call it the UI 10/5 system(Fig. 6).
Assuming that all the UI 10/10 positions have been set properly,we set 10/5 positions on the central sagittal reference curve alongLPA (T9), Cz, and RPA (T10) by 5% increments. Similarly, we set10/5 positions on the central coronal reference curve along Nz, Cz,and Iz by 5% increments. For the 10% axial reference curve, sincethe lengths of its quarterly portions are different, we worked oneach quarter and set 10/5 positions by one tenth increments. Next,we set a coronal reference curve by selecting two corresponding10% axial points and one central sagittal reference point to definethe coronal reference curve (e.g., AFF7, AFFz, and AFF8 for theAFF coronal reference curve) so that the three points are on thesame plane. Since the lengths of the hemispheric portions of thecurves may not be the same, we worked on each hemisphereseparately, placing 10/5 positions by one eighth increments.However, to avoid over-crowded positioning on AFp and POOcoronal reference curves, we worked on each hemisphere to place10/5 positions by one fourth increments as in OostenveldTs 10/5system. For the 0% axial reference curve, since the lengths of its
Fig. 6. The UI 10/5 system. Total number of points is 329 including 12 points, likel10/20 positions, gray open circles indicate additional positions introduced in the UIbe set effectively on a scalp when neighboring positions are not allowed to overlap:red, an additional 50 points preserving only right/left symmetry are in blue, and 6 mThe rest of the UI 10/5 positions (76 points), which interfered with neighboring p
quarterly portions are different, we worked on each quarter and set10/5 positions by one tenth increments. Finally, we set 5%reference curves on the left hemisphere so that NFpz, T9h, and OIzare on the same plane and did the same on the right hemisphere,using NFpz, T10h, and OIz. As in the cases of 0% and 10% axialreference curves, the lengths of the quarterly portions of the 5%axial reference curve are different. Thus, we work on each quarterand set 10/5 positions by one tenth increments. Consequently, weestimated a total of 329 UI 10/5 standard positions and describedtheir variability on the MNI space (Fig. 7).
Validity of 10/20, 10/10, and 10/5 systems
Next, we examined whether landmark positions of 10/20, 10/10,and 10/5 systems could be resolved from neighboring positions. Wemeasured the center-to-center distance between two neighboringpositions (D) and compared it to the sum of the standard deviationsof the two positions (SD1+SD2). We regarded the two positions asseparate when the distance between the two neighboring positionswas larger (i.e., D>SD1+SD2).
y lying on the eyes (shown in gray italics). Black open circles indicate the UI10/10 system. Colored positions (red, blue, and green) are the points that can185 points preserving both anterior/posterior and right/left symmetries are inore points that are just separated from the neighboring positions are in green.ositions, are indicated in gray.
Fig. 7. Locations of the UI 10/5 standard positions (329) in MNI space andtheir spatial variability. MNI coordinates are also shown. The centers of thecircles represent the most likely locations of MNI coordinates for 10/20standard positions. The edges represent the boundaries defined by standarddeviation. Black circles indicate the UI 10/20 positions (19), gray circlesindicate the additional positions in the UI 10/10 system (62). The rest of theUI 10/5 positions (248) are indicated in white. The asterisks and sharpsindicate positions that can be on the eyes, and on or beneath the ear lobes,respectively. (A) Frontal view. (B) Occipital view. (C) Left temporal view.(D) Right temporal view. (E) Top view.
1609V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
A given 10/10 system provides backward compatibility to thecorresponding 10/20 system for at least 19 standard positions.Thus, we extracted them from various branches of the 10/10systems presented above. In all cases, the 10/20 standard positionswere separated from each other.
We extended this analysis to 10/10 systems. First of all, in theUI 10/10 system, 81 positions were all separated from each otheron a scalp in MNI space. Indeed, in all the other branches, 10/10positions were separated from each other except for a marginallevel of overlapping between PO3 and O1 as determined by ashortest-distant search method for coronal reference curve setting(Fig. 5). If we tolerate these negligible or minor exceptions, we canconclude that the UI 10/10 system provides stable and well-separated landmarks on the scalp.
We further examined whether the UI and Oostenveld’s 10/5positions could be resolved from neighboring positions. Sincethere were obviously overlapping positions, we eliminated the
more subordinate positions in the following order of preference:10/20, 10/10, and 10/5 positions. Furthermore, we excluded 12positions that interfere with the eyes.
For Oostenveld’s 10/5 system, 241 points survived as distinctpositions. In general, anterior positions were more clearlyseparated than posterior positions (Supplementary material 5). Inorder to set a maximum balanced separation to achieve right/leftsymmetry, we further eliminated six positions (Supplementarymaterial 5). In addition to the right/left symmetry, we furthersought anterior/posterior symmetry. Ultimately, 189 points sur-vived. These positions may serve as a fair criterion for setting upscalp landmarks in Oostenveld’s 10/5 system.
For the UI 10/5 system, the results were similar as forOostenveld’s 10/5 system. As Fig. 6 shows, 241 points survived asdistinct positions. As in the case of Oostenveld’s 10/5 system,anterior positions were more clearly separated than posteriorpositions. After applying the right/left symmetry criteria, 235positions survived (Fig. 6). After further applying the anterior/posterior symmetry criteria, 185 positions survived. Thesepositions may serve as a fair criterion for setting up scalplandmarks for the UI 10/5 system.
Discussion
The aim of the current study was to evaluate the effectivenessof 10/20-derived systems in the light of head-surface-basedpositioning systems. From the time of its invention as a methodto set up EEG electrodes in a balanced reproducible way, the 10/20system has gained importance as a standard method for settinglandmarks over the scalp.
Nevertheless, the current definitions for the 10/20 system and itsderivatives still remain ambiguous, and this reduces the potentialaccuracy of these systems. Ideally, in order to increase accuracy, thecurrent definitions should be revised to providemore detailedmethodsfor setting landmarks. However, in practice, it takes time to realizesuch standardizations. Even after standardization, it would take timefor the new methods to become widely used. For example, thetransition from Talairach to MNI stereotaxic coordinate systems(reviewed in Brett et al., 2002) is still in process. It is possible, ofcourse, that this transition may never be fully completed and thatresearchers will ultimately opt for the coexistence of the two systems.
Therefore, we have chosen to present methods to probabil-istically describe major branches of 10/20-derived systems. Thesevariations may have developed to satisfy the individual needs ofexperimenters and clinicians and would therefore be useful in somesituations. There were inter-system variabilities between 10/10positions defined in different methods. Specifically, choice ofhemispheric or anterior/posterior divisions and of coronal referencecurves causes non-negligible differences in the locations of 10/10positions. These observations clearly demonstrate the need toclarify which branch of 10/20-derived systems is used to set scalplandmarks in order to describe their locations explicitly. On theother hand, intra-system variabilities were at similar levels. Fromthese observations, we conclude that, as long as a detailed rule for aparticular method is provided, it will yield precise landmarks. Inaddition, we should stress here that we do not intend to judgewhich system is superior but to demonstrate that landmarks set byany system can be probabilistically described.
In the best case scenario, with a clear description of the rules ofmeasurement, the 10/5 system can set as many as 329 positions ona scalp (Oostenveld and Praamstra, 2001). Full use of these
1610 V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
positions may be useful for EEG experiments, but such highdensity setting may result in overestimations for relative headsurface positioning. When two neighboring positions are notallowed to overlap in terms of standard deviations, the number ofeffective 10/5 positions was reduced to 241 for Oostenveld’s andUI 10/5 systems. Moreover, when they were balanced out foranterior/posterior and right/left symmetries, the number of effective10/5 positions was reduced to 189 for Oostenveld’s system and 185for the UI 10/5 system. However, it would not be unrealistic tostate that, with careful measurements, the 10/5 system is capable ofproviding more than 180 distinct landmarks on a scalp.
The criteria for separation used in this study, namely, non-overlapping of SD of neighboring positions, are only roughguidelines. There is the possibility of type I errors, whereinseparable positions are judged as separable (see Supplementary7 for the results of the Monte Carlo simulation for type I errors).
It is worthwhile mentioning that all 10/10 positions remainedintact after the exclusion of overlapping points. In other branchesof 10/10 systems also, most 10/10 positions, with only a fewmarginal exceptions, were well separated from each other. Thus,landmark setting according to a given 10/10 system can also beconsidered reliable as long as explicit definitions are provided.
The virtue of the current study would be most appreciated in thecontext of a cross-modal approach, which is now common in theneuroimaging community. Since there is no single perfect modalityfor assessing human brain function, ideally, data from differentmodalities should be integrated into a single common platform. Inpursuit of a common arena for inter-modal assessment, there hasbeen a movement described as a probabilistic approach toexpressing all functional brain data as entries in a brain atlas thatexpands into space and time (Abbott, 2003; Mazziotta et al., 2000;Mazziotta et al., 2001a,b; Toga and Thompson, 2001). Theprobabilistic atlas per se has not been realized, but its philosophyhas become widespread so that the essential concept for thisintegrative approach has already been realized. It is now commonpractice to present tomographic imaging data in stereotaxicstandard coordinate systems such as Talairach or MNI coordinates(Talairach and Tournoux, 1988; Collins et al., 1994; reviewed inBrett et al., 2002). With respect to expressing the transcranialbrain mapping data on MNI space, the current study will bebeneficial for three major technical applications. First, the currentdata set will provide finer MNI coordinate estimation for EEGsignal source elucidation (Pascual-Marqui et al., 2002). It is alsoplausible to elucidate the accuracy of the signal source estimationby applying error propagation law or resampling simulations(Singh et al., 2005; Tsuzuki et al., 2006). Second, the data setwill be used as standard landmarks to guide fNIRS probes or aTMS coil on a scalp for reproducible measurements. Thereliability of the selected landmarks can be evaluated by errorinformation. Third, the data set will present more referencelandmark positions to perform probabilistic registration of fNIRSand TMS data to standard stereotaxic space (Okamoto et al.,2004a; Okamoto and Dan, 2005; Singh et al., 2005). In theory,we can probabilistically register transcranial data without MRimages with just four reference points, but inclusion of morereference points selected in a balanced way enhances the accuracyof the probabilistic registration.
The current study may also be viewed as part of retrospectivetrends in the neuroimaging community. In one direction, there areextensive cytoarchitectural and areal parcellation studies beingundertaken to provide probabilistic anatomical basis for the MNI
system (Eickhoff et al., 2005; reviewed in Amunts and Zilles,2001). These works are considered to be a modern implementationof Brodmann’s legendary works to describe the cytoarchitecture ofthe brain (Brodmann, 1908). In a different direction, we areseeking to re-establish Jasper’s work in a modern perspective inorder to create a link between stereotaxic coordinates and relativehead-surface-based positioning systems. Such an approach wouldbe beneficial for creating a tighter link between tomographic andtranscranial brain mapping methods. Powered by old ideas, webelieve that this study will accelerate the movement for cross-model data sharing and integration on a common platform.
Acknowledgments
We thank Ms. Archana K. Singh, Dr. Haruka Dan, and Dr.Masako Okamoto for examination of the manuscript, Ms. AkikoOishi and Ms. Yumiko Shiga for preparation of the manuscript anddata, and Ms. Melissa Nuytten for examination of the manuscript.We appreciate Dr. Ryusuke Kakigi and Dr. Roberto D. Pascual-Marqui for giving us the initial inspiration for the current work.This work is supported by the Industrial Technology ResearchGrant Program in 03A47022 from the New Energy and IndustrialTechnology Development Organization (NEDO) of Japan and theProgram for Promotion of Basic Research Activities for InnovativeBioscience (PROBRAIN).
Appendix A. Supplementary data
Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.neuroimage.2006.09.024.
References
Abbott, A., 2003. Neuroscience: a new atlas of the brain. Nature 424,249–250.
American Electroencephalographic Society, 1994. Guideline thirteen:guidelines for standard electrode position nomenclature. AmericanElectroencephalographic Society. J. Clin. Neurophysiol. 11, 111–113.
Amunts, K., Zilles, K., 2001. Advances in cytoarchitectonic mapping of thehuman cerebral cortex. Neuroimaging Clin. N. Am. 11, 151–169.
Brett, M., Johnsrude, I.S., Owen, A.M., 2002. The problem of functionallocalization in the human brain. Nat. Rev., Neurosci. 3, 243–249.
Brodmann, K., 1908. Beiträge zur histologischen Lokalisation derGroβhirnrinde. VI: Die Cortexgliederung des Menschen. J. Psychol.Neurol. X, 231–246 (in German).
Chatrian, G.E., Lettich, E., Nelson, P.L., 1985. Ten percent electrode systemfor topographic studies of spontaneous and evoked EEG activity. Am. J.EEG Technol. 25, 83–92.
Chatrian, G.E., Lettich, E., Nelson, P.L., 1988. Modified nomenclature forthe “10%" electrode system. J. Clin. Neurphysiol. 5, 183–186.
Collins, D.L., Neelin, P., Peters, T.M., Evans, A.C., 1994. Automatic 3Dintersubject registration of MR volumetric data in standardized Talairachspace. J. Comput. Assist. Tomogr. 18, 192–205.
Eickhoff, S.B., Stephan, K.E., Mohlberg, H., Grefkes, C., Fink, G.R.,Amunts, K., Zilles, K., 2005. A new SPM toolbox for combining pro-babilistic cytoarchitectonic maps and functional imaging data. Neuro-Image 25, 1325–1335.
Herwig, U., Satrapi, P., Schonfeldt-Lecuona, C., 2003. Using theinternational 10–20 EEG system for positioning of transcranial magneticstimulation. Brain Topogr. 16, 95–99.
Jasper, H.H., 1958. The ten–twenty electrode system of the InternationalFederation. Electroencephalogr. Clin. Neurophysiol. 10, 367–380.
1611V. Jurcak et al. / NeuroImage 34 (2007) 1600–1611
Jurcak, V., Okamoto, M., Singh, A., Dan, I., 2005. Virtual 10–20measurement on MR images for inter-modal linking of transcranialand tomographic neuroimaging methods. NeuroImage 26, 1184–1192.
Klem, G.H., Luders, H.O., Jasper, H.H., Elger, C., 1999. The ten–twentyelectrode system of the International Federation. The InternationalFederation of Clinical Neurophysiology. Electroencephalogr. Clin.Neurophysiol., Suppl. 52, 3–6.
Le, J., Lu, M., Pellouchoud, E., Gevins, A., 1998. A rapid method fordetermining standard 10/10 electrode positions for high resolution EEGstudies. Electroencephalogr. Clin. Neurophysiol. 106, 554–558.
Mazziotta, J.C., Toga, A.W., Evans, A.C., Fox, P., Lancaster, J., Woods, R.,2000. A probabilistic approach for mapping the human brain. In: Toga,A.W., Mazziotta, J.C. (Eds.), Brain Mapping: The System. AcademicPress, San Diego, pp. 141–156.
Mazziotta, J., Toga, A., Evans, A., Fox, P., Lancaster, J., Zilles, K.,Woods, R., Paus, T., Simpson, G., Pike, B., Holmes, C., Collins, L.,Thompson, P., MacDonald, D., Iacoboni, M., Schormann, T., Amunts,K., Palomero-Gallagher, N., Geyer, S., Parsons, L., Narr, K., Kabani, N.,Le Goualher, G., Boomsma, D., Cannon, T., Kawashima, R., Mazoyer,B., 2001a. A probabilistic atlas and reference system for the humanbrain: International Consortium for Brain Mapping (ICBM). Philos.Trans. R. Soc. Lond., B Biol. Sci. 356, 1293–1322.
Mazziotta, J., Toga, A., Evans, A., Fox, P., Lancaster, J., Zilles, K., Woods,R., Paus, T., Simpson, G., Pike, B., Holmes, C., Collins, L.,Thompson, P., MacDonald, D., Iacoboni, M., Schormann, T., Amunts,K., Palomero-Gallagher, N., Geyer, S., Parsons, L., Narr, K., Kabani, N.,Le Goualher, G., Feidler, J., Smith, K., Boomsma, D., Hulshoff Pol, H.,Cannon, T., Kawashima, R., Mazoyer, B., 2001b. A four-dimensionalprobabilistic atlas of the human brain. J. Am. Med. Inform. Assoc. 8,401–430.
Nuwer, M.R., 1987. Recording electrode site nomenclature. J. Clin.Neurophysiol. 4, 121–133.
Nuwer, M.R., Comi, G., Emerson, R., Fuglsang-Frederiksen, A., Guerit,J.M., Hinrichs, H., Ikeda, A., Luccas, F.J., Rappelsburger, P., 1998.IFCN standards for digital recording of clinical EEG. International
Federation of Clinical Neurophysiology. Electroencephalogr. Clin.Neurophysiol. 106, 259–261.
Okamoto, M., Dan, I., 2005. Automated cortical projection of head-surfacelocations for transcranial functional brain mapping. NeuroImage 26,18–28.
Okamoto, M., Dan, H., Sakamoto, K., Takeo, K., Shimizu, K., Kohno, S.,Oda, I., Isobe, S., Suzuki, T., Kohyama, K., Dan, I., 2004a. Three-dimensional probabilistic anatomical cranio-cerebral correlation via theinternational 10–20 system oriented for transcranial functional brainmapping. NeuroImage 21, 99–111.
Okamoto, M., Dan, H., Shimizu, K., Takeo, K., Amita, T., Oda, I., Konishi,I., Sakamoto, K., Isobe, S., Suzuki, T., Kohyama, K., Dan, I., 2004b.Multimodal assessment of cortical activation during apple peeling byNIRS and fMRI. NeuroImage 21, 1275–1288.
Oostenveld, R., Praamstra, P., 2001. The five percent electrode system forhigh-resolution EEG and ERP measurements. Clin. Neurophysiol. 112,713–719.
Pascual-Marqui, R.D., Esslen, M., Kochi, K., Lehmann, D., 2002.Functional imaging with low-resolution brain electromagnetic tomo-graphy (LORETA): a review. Methods Find. Exp. Clin. Pharmacol. 24(Suppl. C), 91–95.
Singh, A.K., Okamoto, M., Dan, H., Jurcak, V., Dan, I., 2005. Spatialregistration of multichannel multi-subject fNIRS data to MNI spacewithout MRI. NeuroImage 27, 842–851.
Suarez, E., Viegas, M.D., Adjouadi, M., Barreto, A., 2000. Relating inducedchanges in EEG signals to orientation of visual stimuli using the ESI-256machine. Biomed. Sci. Instrum. 36, 33–38.
Talairach, J., Tournoux, P., 1988. Co-Planar Stereotaxic Atlas of the HumanBrain: 3-Dimensional Proportional System: An Approach to CerebralImaging. Thieme Medical Publishers, New York.
Toga, A.W., Thompson, P.M., 2001. Maps of the brain. Anat. Rec. 265,37–53.
Tsuzuki, D., Jurcak, V., Singh, A.K., Okamoto, M., Watanabe, E., Dan, I.,2006. Virtual spatial registration of stand-alone fNRS data to MNI space.NeuroImage, doi:10.1016/j.neuroimage.2006.10.043.
